An Application Layer Response Optimization Method for AIOps

Abstract

An application layer response optimization method for AIOps, including: S1: Establish the serial number relationship of each sub-module in each layer of AIOps; S2: Analyze the data connection between each layer of AIOps in a specific business environment by accessing the historical database, and generate an initial line ID in combination with the serial numbers of the sub-modules in each layer of AIOps; S3: Based on the ant colony algorithm, calculate the shortest line and other lines of the connection of the sub-modules in each layer of AIOps in a specific business environment, and generate the ant colony algorithm line ID in combination with the serial numbers of the sub-modules in each layer of AIOps; S4: Based on the Pareto analysis method, calculate the optimal solutions of the sub-modules in some layers of AIOps for each line in the ant colony algorithm line ID, and then obtain the optimized line. Through the above technical solution, the path from the data acquisition layer to the application layer is optimized, and the output response speed of the application layer is improved.

Description

Technical Field

[0001] The present invention relates to the field of network optimization technology, and particularly relates to an application layer response optimization method for AIOps. Background Art

[0002] Generally, the customer tolerance for the loading of general web pages, WeChat, and APP pages is within 3 seconds. With the large-scale application of big data and AI, how to ensure that the application layer display can be presented within 3 seconds is a technical bottleneck faced today.

[0003] This application introduces an AIOps application layer response optimization method based on the ant colony algorithm combined with Pareto. Through this method, the series relationship of the horizontal acquisition layer, algorithm layer, component layer, and application layer of AIOps can be understood; vertically, the resource load and configuration of each sub-module and component can be understood, and at the same time, the corresponding resources of the components or modules with high load in the operation results of the artificial intelligence model are optimized and upgraded, so that the path from the data acquisition layer to the application layer in AIOps is optimized, and the output response speed of the application layer is improved. Summary of the Invention

[0004] In view of the deficiencies in the prior art, the present invention provides an application layer response optimization method for AIOps to optimize the path from the data acquisition layer to the application layer and improve the output response speed of the application layer.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] An application layer response optimization method for AIOps includes the following steps:

[0007] S1: Establish the serial number relationship of each sub-module in each layer of AIOps;

[0008] S2: Analyze the data connection between each layer of AIOps in a specific business environment by accessing the historical database, and at the same time generate an initial line ID in combination with the serial numbers of the sub-modules in each layer of AIOps;

[0009] S3: Based on the ant colony algorithm, calculate the shortest line and other lines connecting the sub-modules in each layer of AIOps in a specific business environment, and at the same time generate an ant colony algorithm line ID in combination with the serial numbers of the sub-modules in each layer of AIOps;

[0010] S4: Based on the Pareto analysis method, calculate the optimal solutions of the sub-modules in some layers of AIOps for each line in the ant colony algorithm line ID, and then obtain the optimized line.

[0011] To optimize the above technical solutions, the specific measures taken also include:

[0012] Furthermore, each layer of AIOps includes a data collection layer, a data operation and maintenance middleware platform, an algorithm layer, a component layer, an SRE / Devops layer, and an application layer; each layer contains its corresponding sub-module.

[0013] Furthermore, it further includes step S5:

[0014] For a specific business environment, analyze the data types collected by the data collection module through historical data, and match each type of data with each module in the application layer.

[0015] According to the data types collected by the data collection module, select the corresponding sub-module in the application layer as the priority output to determine the hierarchical priority relationship of the application layer.

[0016] Furthermore, the specific content of step S2 is:

[0017] By accessing the historical database, analyze the data sources in the data collection layer and the data intersection with the application layer, and then judge the processes participated by other layers in AIOps, generate multiple lines from the data collection layer to the application layer, and select the line with the highest frequency of occurrence as the initial line.

[0018] Combine the serial numbers of the sub-modules in each layer of AIOps to generate an initial line ID for the initial line.

[0019] Furthermore, the specific content of step S3 is:

[0020] Build an ant colony algorithm line model based on the ant colony algorithm, and by accessing the historical database, use the three sets of contents of the fields of each sub-module participating in data processing in the data collection layer, the fields of the sub-modules in the application layer associated with each sub-module in the data collection module, and the initial line ID field as input parameters, put them into the ant colony algorithm line model for calculation to obtain the shortest line and other lines between the data collection layer and the application layer.

[0021] Combine the serial numbers of the sub-modules in each layer of AIOps to generate ant colony algorithm line IDs for the shortest line and other lines.

[0022] Furthermore, the specific content of step S4 is:

[0023] Build a line classification model based on the Pareto analysis method, use the initial line ID, the serial numbers of the application layer sub-modules associated with the data collection layer in the specific business, and the ant colony algorithm line ID as input parameters, and calculate the optimal solutions of each sub-module in the data operation and maintenance middleware platform, the algorithm layer, and the component layer through the line classification model.

[0024] Furthermore, obtain the optimized lines between the data collection layer and the application layer for each line in the ant colony algorithm line ID.

[0025] Further, while calculating the optimal solutions of each sub-module in the data operation and maintenance middle platform, algorithm layer, and component layer through the line classification model, the non-optimal solutions of each sub-module in each layer are also obtained; thus, all other lines between the data acquisition layer and the application layer except the optimized line are known.

[0026] When the optimized line does not meet the business requirements, select from all other lines.

[0027] The beneficial effects of the present invention are as follows: This application can comprehensively understand the series relationship of the AIOps horizontal acquisition layer, algorithm layer, component layer, and application layer; longitudinally understand the resource load and configuration of each sub-module and component, and at the same time optimize and upgrade the corresponding resources of the components or modules with high load in the operation results of the artificial intelligence model. Combining the ant colony algorithm and the Pareto analysis method can solve the problem of lack of series connection of the component and sub-module data in the AIOps multi-layer index system, so as to realize the intelligent scheduling of the acquisition data layer and the application layer in a macroscopic sense and improve the output response speed of the application layer. Detailed implementation manners

[0028] Now, the present invention will be further described in detail.

[0029] The main technical solution of this application is as follows:

[0030] It includes: data acquisition, analysis and generation of line module, construction of line classification model and scenario arrangement module by Pareto analysis method, and analysis and matching of Lineid and application layer module to obtain the optimal hierarchical relationship.

[0031] Step 1: Data acquisition, analysis and generation of line module:

[0032] First, access the historical database through the AIOps data analysis and optimization program to analyze and generate the initial line ID (Lineid).

[0033] Secondly, creatively use the ant colony algorithm to construct the [ant colony algorithm line model] and put the three groups of parameters of each acquisition layer module field, [application layer - acquisition layer association mark] field, and [initial line ID (Lineid)] field in the historical database into the model for calculation. The calculation result is the shortest line of the ant colony algorithm and other lines between the current acquisition layer and the application layer. Then, combine the shortest line and other lines to generate the ant colony algorithm line ID (Lineid).

[0034] AIOps Analysis and Optimization Program: Access the historical database. After analyzing the data sources in the acquisition layer and the application layer - associating tags, concatenate the corresponding identification numbers of the steps that have ever intersected, generate multiple lines, and obtain the line with the most occurrences as the initial line ID (Lineid).

[0035] Step 2: Use the Pareto analysis method to construct a line classification model and a scenario orchestration module:

[0036] Use the Pareto analysis method (Pareto) to construct a [Line Classification Model], classify the calculation results of the AIOps operation and maintenance data center, algorithm layer, and component layer ant colony algorithm to obtain the Pareto optimal solution. At the same time, obtain all the lines classified in each layer.

[0037] Step 3: Analyze and match Lineid with the application layer module to obtain the optimal hierarchical relationship: Parse (Lineid) and extract the sub-module analysis and matching of the acquisition layer and the application layer, that is: the module in the acquisition layer to the module in the application layer is the optimal application and display of the AIOps acquisition data. Then, bind the two and generate an optimal hierarchical relationship identifier, which is updated to the database. The format of the optimal hierarchical relationship identifier: Acquisition layer identifier @ Application layer identifier = 1-1 @ 1. For unmatched (Lineid), parse other line matches from best to worst according to the sorting. If still unable to match, use the previous associated line. Thus, complete the entire process of associating the optimal line with the optimal relationship in the application layer.

[0038] The specific solution process is as follows:

[0039] Step 1: Data acquisition, analysis, and generation of line module:

[0040] First, access the historical database through the AIOps data analysis and optimization program to analyze and generate the initial line ID (Lineid).

[0041] (The serial number relationships of each sub-module in AIOps are as follows:

[0042] The data acquisition layer includes: 1-1 = Environmental monitoring, 1-2 = Network monitoring, 1-3 = Host monitoring, 1-4 = System monitoring, 1-5 = Security monitoring, 1-6 = Cloud resources, 1-7 = Application monitoring;

[0043] The data operation and maintenance center: 2-1 = Data cleaning, 2-2 = Data filtering, 2-3 = Data standardization, 2-4 = Index operation, 2-5 = Anomaly marking, 2-6 = Data warehouse CMDB: 2-7 = CI configuration item management, 2-8 = CI relationship management, 2-9 = Business model topology, 2-10 = Dictionary and rule management;

[0044] Algorithm layer: 3-1 = Multi-objective optimization, 3-2 = NLP log algorithm, 3-3 = Neural network, 3-4 = Hybrid graph model, 3-4 = Metric, 3-5 = Trace;

[0045] Component layer: 4-1 = Intelligent analysis: Sub-components include (4-1-1 Association analysis, 4-1-2 Routing analysis, 4-1-3 Log analysis, 4-1-4 Resource analysis);

[0046] 4-2 = Service management: Sub-components include (4-2-1 = Service catalog, 4-2-2 = Work order management, 4-2-3 = Personnel management, 4-2-4 Workplace management);

[0047] 4-3 = Automated operation: Sub-components include (4-3-1 = Automatic inspection, 4-3-2 = Batch operation, 4-3-3 = Emergency handling, 4-3-4 = Customization);

[0048] 4-4 = Knowledge base: Sub-components include (4-4-1 = Fault handling knowledge base, 4-4-2 = Engineer experience base);

[0049] 4-5 = 3D visualization engine: Sub-components include (4-5-1 = Graphics rendering, 4-5-2 = Application interface);

[0050] SRE / Devops layer: 5-1 = Atomic capabilities, 5-2 = Capacity management, 5-3 = Code management, 5-4 = Automatic release, 5-5 = API / SDK;

[0051] Unique serial number of the application layer database: 1 = Large screen real-time monitoring, 2 = Management cockpit, 3 = Fault precise positioning, 4 = Automatic fault handling, 5 = Fault prediction, 6 = Digital twin, 7 = Alarm intelligent analysis, 8 = Intelligent dispatch, 9 = Intelligent report, 10 = Mobile office 11 = App;

[0052] # represents the line separator, and @ represents the separator for levels and sub-modules).

[0053] Secondly, creatively adopt the ant colony algorithm to construct the

Ant colony algorithm line model

Application layer - Acquisition layer association mark

Initial line ID (Lineid)

[0054] AIOps analysis and optimization program: Access the historical database, concatenate the corresponding identification numbers of the steps that used to have intersections after analyzing the data sources of the acquisition layer and the application layer - association marks, generate multiple lines, and obtain the line with the most occurrences as the initial line ID (Lineid).

[0055] Table 1 Introduction to the optimized structure of the historical database (specific implementation examples are given for the first environmental monitoring)

[0056]

[0057]

[0058]

Ant Colony Algorithm Model

[0059] (Ant Colony Algorithm) Core formula and description:

[0060]

[0061] The probability that the kth ant in the tth generation of ants chooses to go to Guandong or Xikou, that is, the probability that ant k chooses to go from ij; a: the importance of pheromone; β: the relative importance of the heuristic factor; n ij : Heuristic factor; J k (i): The city that ant k can choose in the current period (note: each city can only be visited once); And d ij Represents the distance of ij.

[0062] Step 2: Use Pareto analysis to build a route classification model and scenario arrangement module:

[0063] The Pareto analysis method (Pareto) is used to build a [Line Classification Model], and the calculation results of the ant colony algorithm at the AIOps operation and maintenance data center, algorithm layer, and component layer are classified to obtain the Pareto optimal solution. At the same time, all the lines classified at each layer are obtained.

[0064] Build a route classification model

[0065] Pareto analysis (Pareto) formula: minf(x) = (f 1 (x), …, f p (x) T ,

[0066] The variable feasible domain is S, and the corresponding target feasible domain is Z = f(S); given a feasible point x*∈S, we have If f(x*)<f(x), then x* is called the absolute optimal solution to the multi-objective programming problem. If there is no x∈S such that f(x)<f(x*), then x* is called an effective solution to the objective programming problem. An effective solution to a multi-objective programming problem is also called a Pareto optimal solution.

[0067] Detailed description of Pareto optimal solution and all lines of each classification layer:

[0068] S1. Use the [Line Classification Model] to perform computational analysis on the sub-modules in the operation and maintenance data to obtain the optimal operation and maintenance data center line (Pareto optimal solution) and all line data.

[0069] S2. Use the [Line Classification Model] to perform computational analysis on each model in the algorithm layer to obtain the algorithm layer (Pareto optimal solution) line and all line data. Thus, when the data requires the support of the algorithm layer, the algorithm line selection with the best computational efficiency can be quickly provided. If the (Pareto optimal solution) line does not meet the current business, select and adjust through all lines.

[0070] S3. Use the [Line Classification Model] to perform computational analysis on each component in the component layer to obtain the component layer (Pareto optimal solution) line and all line data.

[0071] The specific usage parameters are as follows:

[0072] Parameter 1: Initial line ID (Lineid) # line separator, @ hierarchy and sub-hierarchy separator, separated by commas for multiple;

[0073] Format: Acquisition layer 1-1@Mid-tier 2-2@Algorithm layer@Component layer@SRE / Devops layer@Application layer 1,11.

[0074] Parameter 2: Mark for the application layer to associate with the environmental monitoring of the acquisition layer;

[0075] Format: (1,3,4,5,6,7,8,9,11);

[0076] Unique serial number of the application layer database: 1 = Real-time large screen monitoring, 2 = Management cockpit, 3 = Fault precise positioning, 4 = Automatic fault handling, 5 = Fault prediction, 6 = Digital twin, 7 = Intelligent alarm analysis, 8 = Intelligent dispatch, 9 = Intelligent report, 10 = Mobile office 11 = App.

[0077] Parameter 3: Ant colony algorithm line ID (Lineid) # line separator, @ hierarchy and sub-hierarchy separator, separated by commas for multiple;

[0078] Format: 1-1@2-1@3-1@4-3-4@1,7#1-1@2-2@1,11.

[0079] In addition, the specific descriptions of the six subclasses (Pareto optimal solution) lines in the component layer are as follows:

[0080] The component layer is classified and arranged according to scenarios into six categories, and six sub - classes (Pareto optimal solutions) of the component layer are obtained by executing the AIOps control and scheduling program. At the same time, the (Pareto optimal solution) lines of other AIOps layers are associated, and multiple line IDs (Lineid) are generated, separated by commas.

[0081] Functions of the AIOps analysis and optimization program:

[0082] I. The program accesses the database component layer, uses the six - category classification as the retrieval condition, obtains three groups of parameters, puts them into the [Line Classification Model], calculates six sub - classes (Pareto optimal solutions) of lines and other lines, and generates multiple line IDs (Lineid), separated by commas.

[0083] II. Match the (Lineid) with the sub - modules of the application layer. For the (Pareto optimal solution) lines of each level of AIOps and other lines, if the line combination contains the component layer, first associate with the (Pareto optimal solution) lines of the component layer, then associate with other lines of the component layer to obtain the optimal line and update the line (Lineid).

[0084] The component layer is classified and arranged according to scenarios into six categories:

[0085] 1. Intelligent analysis: Sub - components include (association analysis, routing analysis, log analysis, resource analysis)

[0086] 2. Service management: Sub - components include (service catalog, work order management, personnel management, workplace management)

[0087] 3. Automated operation: Sub - components include (automatic inspection, batch operation, emergency handling, customization)

[0088] 4. Knowledge base: Sub - components include (fault handling knowledge base, engineer experience base)

[0089] 5. 3D visualization engine: Sub - components include (graphic rendering, application interface)

[0090] 6. SRE / Devops: Sub - components include (atomic capabilities, capacity management, code management, automatic release, API / SDK).

[0091] Step 3. Analyze and match Lineid with application layer modules to obtain the optimal hierarchical relationship: Parse (Lineid) and extract the sub-module analysis and matching between the acquisition layer and the application layer, that is: the module in the acquisition layer to the module in the application layer is the optimal application and display for AIOps to collect data. Then, bind the two and generate an optimal hierarchical relationship identifier, which is updated to the database. The format of the optimal hierarchical relationship identifier: Acquisition layer identifier @ Application layer identifier = 1-1 @ 1. If it does not match (Lineid), parse and match other lines in descending order, and if it still cannot be matched, use the previous associated line. Thus, complete all processes of associating the optimal line with the optimal relationship in the application layer.

[0092] Specific description of the optimal hierarchical relationship and generation of the optimal relationship identifier is as follows:

[0093] First, obtain the tags of the environmental monitoring in the application layer - associated with the acquisition layer in the database [1, 3, 4, 5, 6, 7, 8, 9, 11]. These numbers represent the associated relationships between the environmental monitoring modules in the acquisition layer and those application layer modules;

[0094] Unique serial number of the application layer in the database: 1 = Real-time monitoring of large screens, 2 = Management cockpit, 3 = Precise fault location, 4 = Automatic fault handling, 5 = Fault prediction, 6 = Digital twin, 7 = Intelligent alarm analysis, 8 = Intelligent dispatching, 9 = Intelligent report, 10 = Mobile office, 11 = App;

[0095] Secondly, through historical data analysis, find out which type of data has the largest quantity of interaction data between the application layer module and the environmental monitoring module in the acquisition layer. Then, bind the application layer module corresponding to this type of data and set it as [optimal hierarchy].

[0096] Specific description: Divide the database data into three parts according to types: arithmetic data, real-time data, and analytical data:

[0097] 1. Arithmetic data: mainly used for result display in (1 = Real-time monitoring of large screens, 9 = Intelligent report);

[0098] 2. Real-time data: mainly used for display in the most intuitive 3D for users, monitoring large screens, and alarm WeChat official accounts (1 = Real-time monitoring of large screens, 11 = App);

[0099] 3. Analytical data: After the knowledge base data analysis, it responds through the customer service robot. (2 = Management cockpit, 3 = Precise fault location, 4 = Automatic fault handling, 5 = Fault prediction, 6 = Digital twin, 7 = Intelligent alarm analysis);

[0100] Historical data extraction responds to applications such as artificial voice large screens, WeChat, etc. (7 = Intelligent alarm analysis, 11 = App, 3 = Precise fault location).

[0101] Table 2 Introduction to the Optimal Relationship at the Historical Database Level (Taking Environmental Monitoring as an Example)

[0102]

[0103] It should be added that: In the present application, the AIOps platform is creatively combined with the ant colony algorithm and the Pareto analysis method model technology, highlighting the position of artificial intelligence in intelligent data analysis and optimization of the AIOps platform. Step 1, Data Acquisition and Analysis Module: The ant colony algorithm is creatively used to construct an initial optimal [Ant Colony Algorithm Route Model] to plan the route between the data acquisition layer and the application layer, and the module connection provides the initial route basis. Step 2, Pareto Analysis Method to Construct the Route Model and Scenario Orchestration Module: The Pareto analysis method (Pareto) is used to construct a [Route Classification Model]. The operation and analysis of each sub-module of the operation and maintenance data center, algorithm layer, and component layer are carried out to obtain the Pareto optimal solution route and all route data. If the best algorithm route does not conform to the current business, it is adjusted by selecting from all routes. Finally, by executing the AIOps data analysis and optimization program, multiple line IDs (Lineid) are generated, separated by commas; Step 3, Analysis and Matching of Lineid with the Application Layer Module to Obtain the Optimal Relationship at the Hierarchy: After receiving and parsing the Pareto optimal solution route (Lineid), the application layer matches it with the application layer module. If the match is successful, the optimal relationship identifier at the hierarchy is generated and updated to the database. If the match fails, multiple sub-optimal and lower routes provided by (Lineid) are matched one by one. If the match still fails, the previous associated route is used by default.

[0104] It should be noted that terms such as "up", "down", "left", "right", "front", "back", etc. cited in the invention are only for the convenience of clear narration, rather than to limit the scope of implementation of the present invention. The change or adjustment of their relative relationship, without substantial change in the technical content, should also be regarded as the scope of implementation of the present invention.

[0105] The above is only the preferred implementation mode of the present invention. The protection scope of the present invention is not limited to the above embodiments. All technical solutions within the idea of the present invention belong to the protection scope of the present invention. It should be pointed out that for those of ordinary skill in the art in this technical field, several improvements and refinements made without departing from the principle of the present invention should be regarded as within the protection scope of the present invention.

Claims

1. An application layer response optimization method for AIOps, characterized in that, it includes the following steps: S1: Establish the serial number relationship of each sub-module in each layer of AIOps; each layer of AIOps includes a data collection layer, a data operation and maintenance middle platform, an algorithm layer, a component layer, an SRE / Devops layer, and an application layer; each layer contains its own corresponding sub-module; S2: Analyze the data connection between each layer of AIOps in a specific business environment by accessing the historical database, and at the same time generate an initial line ID by combining the serial numbers of the sub-modules in each layer of AIOps; By accessing the historical database, analyze the data source in the data collection layer and the data intersection with the application layer, and then judge the processes participated by other layers in AIOps, generate multiple lines from the data collection layer to the application layer, and select the line with the highest frequency of occurrence as the initial line; Combine the serial numbers of the sub-modules in each layer of AIOps to generate an initial line ID from the initial line; S3: Based on the ant colony algorithm, calculate the shortest line and other lines of the connection of the sub-modules in each layer of AIOps in a specific business environment, and at the same time generate an ant colony algorithm line ID by combining the serial numbers of the sub-modules in each layer of AIOps; S4: Based on the Pareto analysis method, for each line in the ant colony algorithm line ID, calculate the solutions of the sub-modules in some layers of AIOps, and then obtain the optimized line.

2. The application layer response optimization method for AIOps according to claim 1, characterized in that, it further includes step S5: For a specific business environment, analyze the data types collected by the data collection module through historical data, and match each data type with each module in the application layer; According to the data types collected by the data collection module, select the corresponding sub-module in the application layer as the priority output, so as to determine the hierarchical priority relationship of the application layer.

3. The application layer response optimization method for AIOps according to claim 1, characterized in that, The specific content of step S3 is: Based on the ant colony algorithm, construct an ant colony algorithm line model, and by accessing the historical database, use the three sets of contents of the sub-module fields participating in data processing in the data collection layer, the sub-module fields in the application layer associated with each sub-module in the data collection module, and the initial line ID field as input parameters, put them into the ant colony algorithm line model for calculation, and obtain the shortest line and other lines between the data collection layer and the application layer; Combine the serial numbers of the sub-modules in each layer of AIOps to generate an ant colony algorithm line ID from the shortest line and other lines.

4. The application layer response optimization method for AIOps according to claim 3, characterized in that, The specific content of step S4 is: Based on the Pareto analysis method, construct a line classification model, use the initial line ID, the serial numbers of the application layer sub-modules associated with the data collection layer in the specific business, and the ant colony algorithm line ID as input parameters, and calculate the solutions of each sub-module in the data operation and maintenance middle platform, the algorithm layer, and the component layer through the line classification model; Furthermore, obtain the optimized line between the data collection layer and the application layer for each line in the ant colony algorithm line ID.

5. An application layer response optimization method for AIOps according to claim 4, characterized in that, while calculating the solutions of each sub-module in the data operation and maintenance middleware, algorithm layer, and component layer through the line classification model, the non-optimal solutions of each sub-module in each layer are also obtained; furthermore, all other lines between the data acquisition layer and the application layer except the optimized line are known; when the optimized line does not meet the business requirements, select from all other lines.

Images